Reinforcement Learning for Smart Charging of Electric Buses in Smart Grid

In recent years, the environmental issues caused by using conventional energy resources, such as gasoline and diesel, become more and more serious. One promising solution to these issues is the electrification of public transit by replacing the internal combustion engine buses with electric buses (EBs). However, due to the degradation of EB batteries, the optimization of EB charging schedules during operating time is still challenging for the public transit service providers. This challenge is further complicated by the randomnesses of traffic and road conditions. In this paper, the problem of optimizing EB charging schedules is formulated as a Markov decision process, based on the battery degradation model of EBs and the information available via vehicular communication networks in smart grid. A double Q-leaning algorithm is used to optimize the charging schedules by minimizing the battery degradation cost of EBs. The performance of the proposed algorithm is evaluated by comparing with existing algorithms based on the real data of EB mobility and energy consumption collected from St. Albert Transit, AB, Canada.